Organizational playbooks for people and AI agents
Protagon Prime turns your organization's mission, know-how and ways of working into playbooks that people and AI agents execute together: repeatable, inspectable and governable.
↓ Legal · Government · Insurance · Projects & ConstructionSee it run
Pick an industry and watch an approved playbook run: agents do the legwork, a person approves at the decision gate, and the case ends in a work product you can review.
The problem
Calling an agent hundreds of times does not, by itself, create an organizational playbook. The knowledge lives in prompts, the decisions live in chat threads, and every new case starts from scratch.
The playbook
A .pbk playbook packages everything a recurring operation needs, so people and agents can run it the same way, every time.
The mission, scope, intended result and definition of success.
Relevant expertise, sources, policies and available resources.
Steps, inputs, outputs, dependencies and exceptions.
Human decisions, agent roles, tools and permissions.
Checks, review requirements and the record needed to inspect a case.
From process to playbook
Write down how the work is done today, in plain language. Protagon Prime drafts a playbook, reuses agents and skills you already have, and proposes new ones. Nothing new runs until a person reviews it.
The operating model
The playbook defines the work. People and AI agents execute it in a shared workspace, and every case returns evidence you can review and learn from.
Metrics
Protagon Prime records every step, finding and model call, so each run reports what it cost and what it produced. These are the figures from our own demo runs across four industries.
Legal, government, insurance, projects. Four playbooks, from 3 to 9 steps, each ending in a reviewable work product.
Model cost ranged from $0.03 to $2.72 per run. Every model call is priced and rolled up by step and by case.
Approval gates are part of the run. Where a run took hours, the wall clock includes time spent waiting for a person to approve.
Measured on synthetic demo cases; model cost only, excludes human review time. No client data. Comparisons with manual work elsewhere on this page are illustrative, not measured.
Governance
Governance is part of the playbook, not an add-on. Decide who approves, which models run, and where your data lives.
The playbook pauses at defined decisions until the named role reviews and approves.
Every input, action, output, model call and review decision is recorded for each run.
Choose approved models per step: cloud or local. Agents use approved tools only.
Keep data in Canada, or self-host Protagon Prime in your own cloud or on-premise.
Your cases, documents and playbooks are not used to train models.
Versioned, immutable playbooks, role-based permissions and single sign-on (OIDC).
Ask us for our security and data-handling documentation during your demo.
Principles
Reuse the approved method for the next case. Only the facts change.
Assign each step to the right person or a capable agent with approved tools.
See every input, action, output, model call and review decision.
Versioned playbooks, human decision gates, role-based permissions and SSO.
Each run writes a reflection that proposes the next, better version.
Who it is for
Protagon Prime is not for one-off questions. It fits any team whose core operation runs again and again, draws on specialist knowledge, and needs someone to sign off on the result.
A strong fit when the work is
Protagon Prime records what happened in each run (quality, time, cost and scale), so improvements are measured, not guessed.
Getting started
Pick one operation that matters to your mission, and its process owner.
Turn your expertise, rules and review points into a playbook.
Run it on approved cases and measure against the scorecard.
Publish the version that works and reuse it across the organization.
Tell us about one recurring operation. We will show you what its playbook could look like.
Book a demo